1. Morphological Ergativity

Ch. 11: Syntacticergativity1

Chapter 11
Syntacticergativity

1. Morphological ergativity

As we saw in the preceding chapter, the subject of a German clause normally takes nominative case. This is true whether the clause is transitive, as in (1a), or intransitive, as in (1b). Direct objects take a different case marker, namely the accusative (1c).

(1) a.DerHundhatmichgebissen.
the(masc.nom)doghasme(acc)bitten
‘The dog bit me.’

b.DerHundschläft.
the(masc.nom)dogsleeps
‘The dog is sleeping.’

c.IchsahdenHund.
I(nom)sawthe(masc.acc)dog
‘I saw the dog.’

The case-marking pattern for masculine singular definite subject and object NPs in German is displayed in table (2). We are primarily interested in the fact that intransitive subjects take the same marker (der) as transitive agents, while transitive patients take a distinct marker (den). This kind of pattern is cross-linguistically the most common among case-marking languages. It is referred to as a Nominative-Accusative (or sometimes just “accusative”) system, because these are the names which are normally used for the two cases involved.

(2) / intransitive / Subject: der
transitive / Agent: der / Patient: den

A different case-marking pattern, which is found in a smaller but still significant number of languages, is illustrated in the following data:

(3) Walmatjari (Australia; adapted from J. Healey 1990; data from Hudson 1978)

a.parripalaparni.
boyauxrun
‘The boy ran.’

b.mangapalaparni.
girlauxrun
‘The girl ran.’

c.wirlkapalaparni.
lizardauxrun
‘The lizard ran.’

d.parripapinyamanga-ngu.
boyauxhitgirl
‘The girl hit the boy.’

e.wirlkapanyanyaparri-ngu.
lizardauxseeboy
‘The boy saw the lizard.’

f.ngapapanyanyamanga-ngu.
waterauxseegirl
‘The girl saw the water.’

These examples show that in Walmatjari, patients of transitive clauses get the same case marking as subjects of intransitive clauses, namely Ø. Agents of transitive clauses get a special case marker (ngu). This pattern is displayed in (4).

(4) / intransitive / Subject: -Ø
transitive / Agent: -ngu / Patient: -Ø

This kind of case-marking pattern, in which transitive agents take a unique marker, is referred to as an ergative system. The case marker used for transitive agents is called ergative case, while the case marker (often zero, as in Walmatjari) used for transitive patients and intransitive subjects is called absolutive case. The ergative and accusative patterns are compared in the following diagram, using “S” for the subject of an intransitive clause, “A” for the agent of a transitive clause, and “P” for the patient of a transitive clause.

(5)

2. Ergativity and grammatical relations

In an ergative system like (4), the relationship between case marking and grammatical relations is not immediately obvious. In an intransitive clause like (3a-c), the absolutive argument is clearly the subject. But in a transitive clause like (3d-f), this same case is used for the patient, while the agent takes a different case marker (ergative). This pattern raises an obvious question: which argument of the transitive clause is the grammatical subject, the ergative agent or the absolutive patient? The answer is: it depends. We cannot tell without more information about the syntactic behavior of each argument.

As we emphasized in the preceding chapter, grammatical relations like SUBJ and OBJ are syntactic concepts. The subject of a sentence must be identified on the basis of its syntactic properties.[1] Case marking, on the other hand, is a purely morphological device. Case assignment may be determined on the basis of grammatical relations or semantic roles, or some combination of the two. In addition, we have seen that case assignment can be determined by a special feature in the lexical entry of a particular verb or preposition. Thus we cannot identify the grammatical relations of the arguments in a clause simply by looking at the case-marking patterns.

It turns out that ergative languages differ in their syntactic properties. Warlpiri is another Australian language (related to Walmatjari) with an ergative case-marking pattern, as the following examples demonstrate:

(6) Warlpiri (Simpson, 1991:155 ff)

a.Ngaju-Øka-rnaparnka-mi.
I-abspres-1sg.subjrun-nonpast
‘I am running.’

b.Ngarrka-Øka-Øparnka-mi.
man-abspres-3sg.subjrun-nonpast
‘A man is running.’

c.Ngajulu-rluka-rna-Øngarrka-Ønya-nyi.
I-ergpres-1sg.subj-3sg.objman-abssee-nonpast
‘I see the man.’

d.Ngarrka-ngkuka-Ø-jungaju-Ønya-nyi.[2]
man-ergpres-3sg.subj-1sg.objme-abssee-nonpast
‘A/the man sees me.’

On the other hand, the pattern of verbal agreement follows a NOM-ACC pattern. As the examples in (6) illustrate, the same subject-agreement clitics are used for both the absolutive argument of an intransitive clause and the ergative argument of a transitive clause, e.g. first person singular rna in (6a) and (6c). A distinct object-agreement clitic is used for the patient of a transitive clause, e.g. ju in (6d). Where more than one overt agreement clitic occurs, as in (7), the subject agreement marker normally comes first:

(7) Warlpiri (Simpson, 1991:161)

a.luwa-rnu-rna-ngku
shoot-past-1sg.subj-2sg.obj
‘I shot you.’

b.luwa-rnu-npa-ju
shoot-past-2sg.subj-1sg.obj
‘You shot me.’

So the morphological evidence is mixed: case marking is ergative, while agreement is accusative. We will need to appeal to syntactic evidence to determine which argument is the subject.

The clearest tests for grammatical subjecthood in Warlpiri involve the control relation (see Hale 1982, 1983; Bresnan and Simpson 1983).[3] For example, in adverbial clauses formed with the complementizer =karra, both controller and controllee must be the grammatical subjects of their respective clauses. As the examples in (8) illustrate, the controller may be either the absolutive argument of an intransitive clause (8a) or the ergative argument of a transitive clause (8b-c). Similarly, the controllee may be either the absolutive argument of an intransitive clause (8b) or the ergative argument of a transitive clause (8a,c,d). So this control pattern groups transitive agents with intransitive subjects. In spite of the difference in case marking (absolutive vs. ergative), both types of arguments bear the same grammatical relation, namely subject.

(8) Warlpiri (Hale, 1982; Simpson, 1991:310)

a.Ngarrkakawirnpirli-mi,karlijarnti-rninja=karra.
man(abs)preswhistle-nonpastboomerang(abs)trim-purp=comp
‘The man is whistling, while trimming the boomerang.’

b.Wati-ngkimarlunya-ngujarnti-rninja=karra-rlu.[4]
man-ergkangaroo(abs)see-pastrun-purp=comp-erg
‘The man saw the kangaroo while he (the man) was running.’

c.Ngarrka-ngkukapurlapayunpa-rni,
man-ergprescorroboree(abs)sing-nonpast
karlijarnti-rninja=karra-rlu.
boomerangtrim-purp=comp-erg
‘The man is singing a corroboree, while trimming the boomerang.’

d.Napurrulaka-jungaju-kuwangka-mi,
Napurrula(abs)pres-1sg.objme-datspeak-nonpast
ngurlukipi-rninja=karra.
seedwinnow-purp=comp
‘Napurrula is speaking to me, while winnowing seed.’

Example (8d) illustrates the fact that it is grammatical relations, and not case marking, which determines the control pattern in this construction. The verb wangka ‘speak’ is one of a restricted class of verbs that takes a special, lexically determined case-marking pattern, in this instance an absolutive agent and a dative goal. As (8d) shows, the absolutive agent of this verb can also function as the controller of the =karra clause, because it is the grammatical subject of its clause.

Hale (1982, 1983), Simpson (1991), Laughren (1989, 1992) and Levin (1983) discuss several other kinds of evidence which can be used to identify grammatical subjects in Warlpiri. In each case, the evidence identifies the agent of a transitive clause as the subject, and not the patient. In other words, the ergative case-marking pattern in Warlpiri is purely morphological. The assignment of grammatical relations follows a nominative-accusative pattern: transitive agents (A) and intransitive absolutive arguments (S) are both realized as grammatical subjects; transitive patients (P) bear a distinct relation, namely object.

In some other languages, however, syntactic properties themselves follow an ergative pattern: subjecthood tests pick out the patient of a transitive verb and the single argument of an intransitive verb. The best known example of this type is Dyirbal, another language of Australia.

3. Subjecthood tests in Dyirbal

The term syntacticergativity refers to a situation in which the syntactic system of a language, and in particular the properties which provide tests for subjecthood, follow an ergative pattern. That is, in a syntactically ergative language like Dyirbal, the grammatical subject of a basic transitive clause like (9) is in fact the patient (‘the woman’). This statement may seem hard to accept, because such a pattern is quite unfamiliar to speakers of most other languages. Let us therefore examine some of the syntactic tests which support this claim.[5]

(9)balan dyugumbilbagul yaagubalga-n.
det-woman-absdet-man-erghit-real[6]
‘The man hit the woman.’ (Dixon 1972:130)

Morphologically, Dyirbal exhibits a “split-ergative” case marking system. This term means that some NPs follow the ergative pattern, while others do not. In Dyirbal, 1st and 2nd person pronouns follow the NOM-ACC pattern, while all other NPs follow the ergative pattern. The basic distribution of case markers is summarized in (10); but note that these markers have a number of different allomorphs.

(10) Dyirbal case markers(Dixon 1979:87)

1st & 2nd pers. pro. / 3rd person pronouns / common nouns /
proper names
intransitive subjects / -Ø / -Ø / -Ø
transitive agents / -Ø / -gu / -gu
transitive patients / -nya / -Ø / -Ø

Dyirbal has four genders, or noun classes. The gender of a common noun is reflected in the form of the determiner, which also indicates the case of the NP and the degree of proximity to the speaker (‘here’, ‘there but visible’, or ‘there and not visible’). The various case and gender forms of the medial demonstrative, which signals ‘there but visible’, are listed in (11). These are the determiners used in the examples below.

(11) Dyirbal medial demonstratives (Dixon 1972:44-47, 306-311)

absolutive / erg/instr / dative / genitive
Class I / bayi / bagul / bagul / baul
Class II / balan / bagun / bagun / baun
Class III / balam / bagum / bagum / —
Class IV / bala / bagu / bagu / bau

However, the difference in case marking between 1st and 2nd person pronouns vs. all other NPs is irrelevant to the syntax. In all of the constructions discussed below, subjects of intransitive verbs (S) and patients of transitive verbs (P) play a unique role, no matter whether they are morphologically marked for absolutive, nominative or accusative case. In the discussion that follows, we will use the term absolutive argument to refer to the union of these two classes, S and P, regardless of case marking.

3.1 Relativization

In Dyirbal relative clauses, the modifying clause follows the head noun and is marked by adding the suffix  to the verb. Example (12b) shows a relative clause based on the simple intransitive clause in (12a). The relativized function in (12b) corresponds to the intransitive subject, i.e., the absolutive argument of the basic clause. Example (12c) shows a relative clause based on a simple transitive clause meaning ‘the woman saw the man’. The relativized function in (12c) corresponds to the transitive patient, once again the absolutive argument of the basic clause. As example (12d) demonstrates, the relativized verb agrees with the case marking of the head noun, which in this sentence is ergative.

The example in (12e) is ungrammatical. In this case, the relativized argument is the ergative, i.e., the transitive agent, which is not allowed. Relativization of instrumental, dative, or other oblique arguments is also ungrammatical. So relativization in Dyirbal is constrained by the following generalization: the relativized function must correspond to the absolutive argument of the basic clause.

(12) a. bayi yaabani-nyu.
man-abscome.here-past
‘The man came here.’ (Dixon 1972:59)

b.adyabayi yaa[ __bani-u ]bua-n.
1sg.nomman-abs (abs)come-relsee-real
‘I saw the man who came.’ (Foley & Van Valin 1984:112)

c.balan dyugumbil[ adya___bua-u ]nyina-nyu.
woman-abs 1sg-nom(abs)see-relsit-real
‘The woman that I saw is sitting down.’ (Dixon 1972:100)

d.bayi yaabagun dyugumbiu[ ___waynydyi-u-ru ]bua-n.
man-abswoman-erg(abs)go.uphill-erg-relsee-real
‘The woman who was going uphill saw a man.’ (Dixon 1972:101)

e.*bayi yaa[ ___balan dyugumbilbua-u ]bani-nyu.
man-abs (erg)woman-abssee-relcome-real
(for: ‘The man who saw the woman came here.’) (Foley & Van Valin 1984:112)

As we saw in chapter 7, Keenan and Comrie’s cross-linguistic study of relativization predicts that, if only a single class of arguments can be relativized in a particular language, that argument will be the grammatical subject. Thus the fact that only the absolutive argument can be relativized in Dyirbal strongly suggests that it is the grammatical subject.

3.2 Coordination Reduction

When two clauses which contain coreferential arguments are conjoined, the common argument may be deleted in the second clause provided that it is the absolutive argument of both clauses. This is illustrated in the following examples.

In (13a), the absolutive argument (i.e., the transitive patient) of the second clause is missing; it is interpreted as being coreferential with the absolutive argument of the first clause (‘man’). In (13b) the omitted absolutive argument of the second clause can only be interpreted as being coreferential with the absolutive argument (i.e., the patient ‘woman’) of the first clause, and not with the ergative agent ‘man’. But (13c) is ungrammatical, because the omitted argument in the second clause is the agent of a transitive verb, which is not the absolutive argument.

(13) a. bayi yaabani-nyu,bagun dyugumbiu___balga-n.
man-abscome-real,woman-erg(abs)hit-real
‘The man came and was hit by the woman.’ (Dixon 1972:130)

b.balan dyugumbilbagul yaagubalga-n,___bani-nyu.
woman-absman-erghit-real(abs)come-real
‘The woman was hit by the man and came here.’
(*‘The man hit the woman and came here.’) (adapted from Dixon 1972:131)

c.*bayi yaabani-nyu,balan dyugumbil___balga-n
man-abscome-real,woman-abs(erg)hit-real
(for: ‘The man came and hit the woman.’)

So this construction too reveals a syntactic property which is uniquely shared by the intransitive subject (S) and transitive patient (P).

3.3 Non-finite purpose clauses

A non-finite verb form bearing the suffix –i ~ –gu is used in Dyirbal adverbial clauses which express purpose (Dixon 1972:146). In this construction, the absolutive argument of the purpose clause must be omitted (understood), and must be controlled by the absolutive argument of the matrix verb.

In (14a), both controller and controllee are intransitive subjects (S). In (14b), the controllee is an intransitive subject (S). The only possible choice for controller is the matrix patient (P); the matrix agent cannot be interpreted as controller. In (14c) and (14d), both controller and controllee are transitive patients (P). But sentence (14e) is ungrammatical, because the controllee is a transitive agent (A). These examples are especially significant, because there is a strong cross-linguistic tendency for controllees to be grammatical subjects. This evidence provides strong support for the claim that the absolutive argument in Dyirbal is the grammatical subject.

(14) a. bayi yaawalma-nyuwaynydyil-i
man-absget.up-realgo.downhill-purp
‘The man got up in order to go downhill.’ (Dixon 1972:68)

b.balan dyugumbilbagul yaagubalga-nbadyi-gu.
woman-absman-erghit-realfall-purp
‘The man hit the woman to make her/*himself fall.’ (Dixon 1972:68)

c.balam miranybagul yaagudimba-nyuindababil-i.
beans-absman-ergcarry-real2sg.nompeel-purp
‘The man brought beans so that you could scrape them.’ (Dixon 1972:68)

d.bayi yaabagul gubi-gumunda-nbagun dyugumbiubalgal-i.
man-absshaman-erglead-realwoman-erghit-purp
‘The shaman brought the man to be hit by the woman.’ (Dixon 1972:159)

e.*balan dyugumbilbagul yaaguwawu-nbalan nayinbawalmbil-i.
woman-absman-ergfetch-realgirl-absawaken-purp
(for: ‘The man fetched the woman to wake up the girls.’) (Foley & Van Valin 1984:112)

3.4 Equi constructions

Dixon (1979:128-9) describes a pattern found Dyirbal which he labels the “jussive” construction. This construction involves Equi verbs such as giga-l ‘tell to do’. Dixon shows that the controllee must be the absolutive argument of the complement clause. (Notice that the verb of the complement clause bears the purposive suffix.) Thus (15), where the controllee is an intransitive subject (S), and (16), where the controllee is a transitive patient (P), are both fine. Example (17), however, is ungrammatical because the controllee is a transitive agent (A). Once again, in light of the cross-linguistic tendency for controllees to be grammatical subjects, this evidence supports the claim that the absolutive argument in Dyirbal is the grammatical subject.

(15)anayabugiga-nbanagay-gu.
1pl.nommother-abstell-realreturn-purp
‘We told mother to return.’ (Dixon 1979:129)

(16)adyabayi yaagiga-ngubi-gumawal-i
1sg.nomman-abstell-realshaman-ergexamine-purp
‘I told the man to be examined by the doctor.’ (Comrie 1981:112)

(17)*anayabugiga-numabual-i.
we-nommother-abstell-realfather-abssee-purp
(for: ‘We told mother to watch father.’)

3.5 Plural agreement[7]

The suffix -dyay ‘many’ can be added to the verb to indicate that the absolutive argument is plural. It cannot be used to indicate plurality of the transitive agent, as illustrated in (18). Cross-linguistically, if the verb agrees with only one argument in a transitive clause, that argument is normally the subject. So this evidence too is consistent with the hypothesis which we are considering.

(18) a. bayi yaanyinan-dya-nyu.
man-abssit-plural-real
‘Many men are sitting down.’ (Dixon 1972:249)

b.balam miranybagul yaagugundal-dya-nyu.
black.bean-absman-ergput.in-plural-real
‘The man gathered many beans.’
(Not: *‘Many men gathered beans.’) (Dixon 1972:249)

3.6 Scope of adverbial modifiers

The adverbial particle wara ‘wrongly, badly’ always modifies the absolutive argument of its clause. The results of this constraint are clearly seen in sentence (20), where the particle cannot be interpreted as modifying either the transitive agent (‘man’) or the action itself (‘cut’), but only the patient (‘tree’). So this particle creates another context in which S and P share a unique syntactic property which A lacks.

(19)bayi yaawarabani-nyu.
man-abswronglycome-real
‘The wrong man came.’ (Mel’čuk 1979:42)

(20)bala yugubagul yaaguwaranudi-n.
tree-absman-ergwronglycut-real
‘The man cut the wrong tree.’ (Mel’čuk 1979:42; cf. Dixon 1972:118)
(Not: *‘The wrong man cut the tree’; *‘The man cut the tree in the wrong way.’)

To summarize the results of this section, we have examined a wide variety of syntactic constructions in which the absolutive argument plays a unique and pivotal role. Some of the syntactic properties we have discussed are associated cross-linguistically with grammatical subjects, while others seem to be language-specific features. But the special status of the absolutive argument in Dyirbal syntax seems best explained by assuming that Dyirbal is syntactically ergative, i.e., that the absolutive argument is the grammatical subject.